Expanding the Horizons of Synthesizable Materials: A Study of Various Machine Learning Techniques on Materials Synthesizability Prediction
Abstract/Contents
- Abstract
- This study aims to provide insights into better ways of applying machine learning (ML) to predict and discover new materials. Specifically, this research focuses on comparing the performance of different ML techniques that are trained with data from the Materials Project Database and the Inorganic Crystal Structure Database to identify effective ML techniques that can perform materials synthesizability prediction. Machine learning (ML) is a subfield of artificial intelligence that allows computers to learn and improve from data without being explicitly programmed. By using ML, researchers can analyze and interpret large amounts of data related to materials' properties, structure, and performance. This can lead to the discovery of new materials with desirable properties, as well as the optimization of materials synthesis and processing. In this study of synthesizability, we report results on a number of different models for predicting synthesizability under different data imbalance scenarios. By defining the 100% precision-recall curve area be the best performance a perfect ideal model can achieve, our best performance model is a Multi-Layer Perceptron (MLP) model that exhibits a hold-out set precision-recall curve area of 99% when the dataset is balanced. Our results also show that the MLP model is the most resistant to the influence of data imbalance with a hold-out set precision-recall curve area of 98% when the imbalanced dataset has an unlabeled-to-labeled ratio of 4:1 and a hold-out set precision-recall curve area of 90% when the imbalanced dataset has an unlabeled-to-labeled ratio of 16:1.
Description
Type of resource | text |
---|---|
Publication date | June 8, 2023 |
Creators/Contributors
Author | Cai, William |
---|---|
Thesis advisor | Sendek, Austin |
Thesis advisor | Dionne, Jennifer |
Degree granting institution | Stanford University |
Department | Program in Engineering Physics |
Subjects
Subject | Machine learning |
---|---|
Subject | computational material science |
Genre | Text |
Genre | Thesis |
Bibliographic information
Related item |
|
---|---|
DOI | https://doi.org/10.25740/xf925xp9052 |
Location | https://purl.stanford.edu/xf925xp9052 |
Access conditions
- Use and reproduction
- User agrees that, where applicable, content will not be used to identify or to otherwise infringe the privacy or confidentiality rights of individuals. Content distributed via the Stanford Digital Repository may be subject to additional license and use restrictions applied by the depositor.
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 4.0 International license (CC BY-NC).
Preferred citation
- Preferred citation
- Cai, W. (2023). Expanding the Horizons of Synthesizable Materials: A Study of Various Machine Learning Techniques on Materials Synthesizability Prediction. Stanford Digital Repository. Available at https://purl.stanford.edu/xf925xp9052. https://doi.org/10.25740/xf925xp9052.
Collection
Undergraduate Theses, Program in Engineering Physics
View other items in this collection in SearchWorksContact information
Also listed in
Loading usage metrics...